“Mind—Reading” AI Produces a Description of What You

石斯琳

A new research project of Japan can analyze2 a persons brain scans3 and provide a written description of what they have been looking at.

日本一個(gè)新型研究項(xiàng)目能夠分析人類(lèi)大腦的掃描結(jié)果，并得出一份書(shū)面報(bào)告，描述了人們看到的東西。

When a person is looking at a picture， the artificial intelligence is given an fMRI brain scan image. It then produces a written de-scription of what they think the person was viewing. It can offer a description like： “A dog is sitting on the floor in front of an open door.” or “A group of people standing on the beach.” Both of those turn out to be quite accurate4.

當(dāng)受試者看一張圖片時(shí)，人工智能就接受來(lái)自功能性磁共振成像得來(lái)的腦部掃描圖像。緊接著它輸出一份關(guān)于人們認(rèn)為受試者正看見(jiàn)的東西的書(shū)面描述。這個(gè)人工智能得出的書(shū)面描述精準(zhǔn)程度如下：“一只狗坐在敞開(kāi)著的門(mén)前的地板上”或者“一群人站在沙灘上”。經(jīng)確認(rèn)，這兩個(gè)描述都十分精準(zhǔn)。

“Our goal is to understand how the brain represents5 information about the real world，” Ichiro Kobayashi， one of the researchers from Japans Ochanomizu University， told Digital Trends. “Toward such a goal， we showed that our algorithm6 can model and read out perceptual7 contents from human brain activity. And the algorithm do this in the form of sentences. So we modified8 an existing network model. It could generate sentences from images.

小林一郎是來(lái)日本御茶水女子大學(xué)的研究人員之一。他向《電子趨勢(shì)》透露：“我們的目標(biāo)是了解大腦是怎樣呈現(xiàn)現(xiàn)實(shí)世界傳遞給它的信息的。設(shè)定這個(gè)目標(biāo)之后，通過(guò)演示，我們的（計(jì)算機(jī)）算法能夠以句子的形式模擬并讀出大腦活動(dòng)中的感知內(nèi)容。所以我們對(duì)一個(gè)現(xiàn)有的網(wǎng)絡(luò)模型作出修改。它（模型）能夠用句子的形式呈現(xiàn)從圖像中讀到的內(nèi)容。”

However， Kobayashi points out that it still a long way away from real-world application9. “So far， there are not any real-world applica-tions for this，” Kobayashi continued. “Howev-er， in the future， it might be used widely.”

然而，小林一郎指出，要將這個(gè)模型運(yùn)用到現(xiàn)實(shí)中還需要做很多工作。他說(shuō)：“目前該模型還無(wú)法進(jìn)行實(shí)際運(yùn)用。然而，將來(lái)這種技術(shù)可能會(huì)被廣泛使用?！?/p>

As a next step， Shinji Nishimoto， another researcher on the project， told Digital Trends that the team wants to use it to better under-stand how the brain processes10 information.

西本真治作為該項(xiàng)目的另一位研究人員向《電子趨勢(shì)》透露，研究組下一步要利用這個(gè)技術(shù)更進(jìn)一步了解大腦是怎樣加工信息的。

“We would like to understand how the brain works under naturalistic11 conditions，” Nishimoto said. “Toward such a goal， we are planning to find out how various forms of in-formation—vision12， languages， impressions， etc.—are encoded13 in the brain. In particular， we will work on generating descriptions about what a person thinks.”

“我們想要了解大腦在自然條件下是怎么運(yùn)作的。因此，我們計(jì)劃深入研究各種信息（視覺(jué)、語(yǔ)言、印象等等）在大腦中是怎樣加工的。我們特別致力于這個(gè)技術(shù)（以句子形式）對(duì)人類(lèi)大腦中所想的書(shū)面描寫(xiě)?！?/p>

【fMRI （functional magnetic resonance imaging）： 功能性磁共振成像是一種新興的神經(jīng)影像學(xué)方式

Digital Trends：《電子趨勢(shì)》是一個(gè)科技新聞和信息網(wǎng)站，包括科技新聞、評(píng)論、指南、how-to文章，關(guān)于技術(shù)和消費(fèi)電子產(chǎn)品的視頻和播客的描述。】

Notes：

1. description n. 描述

2. analyze vt. 分析；解析

3. scan n. 掃描

4. accurate adj. 準(zhǔn)確的；精確的

5. represent vt. 代表

6. algorithm n. 算法

7. perceptual adj. 感性的；知覺(jué)的

8. modify v. 修改

9. application n. 應(yīng)用；應(yīng)用程序

10. process v. 加工；處理

11. naturalistic adj. 自然的

12. vision n.洞察力

13. encode vt. 編碼